SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2603.07343 · 2026

Learning Concept Bottleneck Models from Mechanistic Explanations

Schrasing Tong, Lalana Kagal, Antonio De Santis, Marco Brambilla

arXiv · PDF · Open in the Atlas

Code that ran

2 of 4 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 2 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2603.07343.json · badge: /api/ran/2603.07343.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
Antonio-Dee/M-CBM community
Papers with Code archive
2 ran · 2 no recorded run
FunctionStatusWhere it lives
elastic_loss_and_acc Ran
ran; our contract draft was wrong, not the code
Antonio-Dee/M-CBM/glm_saga/elasticnet.py
code served (permissive licence) · get_code("db058f8fcbe2bd2f")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a small linear layer
# Input features: 4, Output classes: 3
linear = nn.Linear(4, 3)

# Create small tensors
# Batch size: 2, Input features: 4
X = torch.randn(2, 4)

# Target labels for multinomial (class indices)
y = torch.tensor([0, 2])

# Regularization parameters
lam = 0.1
alpha = 0.5

# Family parameter
family = 'multinomial'

ARGS = {
    'linear': linear,
    'X': X,
    'y': y,
    'lam': lam,
    'alpha': alpha,
    'family': family
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
elastic_loss_and_acc_loader Ran
ran; our contract draft was wrong, not the code
Antonio-Dee/M-CBM/glm_saga/elasticnet.py
code served (permissive licence) · get_code("22bce7009cca4385")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned tuple. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import torch.nn as nn

torch.manual_seed(0)

# Create a simple linear model
linear = nn.Linear(4, 3)

# Create a simple data loader with 2 batches
# Each batch is (X, y) where X has shape (2, 4) and y has shape (2,)
loader = [
    (torch.randn(2, 4), torch.randint(0, 3, (2,))),
    (torch.randn(2, 4), torch.randint(0, 3, (2,)))
]

# Hyperparameters
lam = 0.01
alpha = 0.5

# Optional preprocessor (a simple nn.Module that transforms input)
class SimplePreprocessor(nn.Module):
    def __init__(self):
        super().__init__()
        self.scale = nn.Parameter(torch.ones(4))
    
    def forward(self, x):
        return x * self.scale

preprocess = SimplePreprocessor()

# Family parameter
family = 'multinomial'

ARGS = {
    'linear': linear,
    'loader': loader,
    'lam': lam,
    'alpha': alpha,
    'preprocess': preprocess,
    'family': family
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
glm_saga Not yet run
unverified (harvested, no recorded run)
Antonio-Dee/M-CBM/glm_saga/elasticnet.py
code served (permissive licence) · get_code("900ae5f4b8fae929")
How it ran

no invocation recorded.

train_saga Not yet run
unverified (harvested, no recorded run)
Antonio-Dee/M-CBM/glm_saga/elasticnet.py
code served (permissive licence) · get_code("9c15a2f68ec9a889")
How it ran

no invocation recorded.

Also present in other papers' code

Of 4 samples lifted from this paper's linked repositories, 4 carry a function identity and 0 do not. Of the 4 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 0 later) and 1 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
elastic_loss_and_acc_loader
brick of this paper's assembly: helper
2 other papers
2 dated earlier · 1 only as an assembly brick
2308.03685 (2023-08)
Learning Concise and Descriptive Attributes for Visual Recognition
2405.06855 (2024-05)
Linear Explanations for Individual Neurons
elastic_loss_and_acc
brick of this paper's assembly: helper
1 other paper
1 dated earlier · 1 only as an assembly brick
2308.03685 (2023-08)
Learning Concise and Descriptive Attributes for Visual Recognition
train_saga
brick of this paper's assembly: update_rule
1 other paper
1 dated earlier
2308.03685 (2023-08)
Learning Concise and Descriptive Attributes for Visual Recognition

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Concept Bottleneck Models (CBMs) aim for ante-hoc interpretability by learning a bottleneck layer that predicts interpretable concepts before the decision. Stateof-the-art approaches typically select which concepts to learn via human specification, open knowledge graphs, prompting an LLM, or using general CLIP concepts. However, concepts defined a-priori may not have sufficient predictive power for the task or even be learnable from the available data. As a result, these CBMs often significantly trail their black-box counterpart when controlling for information leakage. To address this, we introduce a novel CBM pipeline named Mechanistic CBM (M-CBM), which builds the bottleneck directly from a black-box model's own learned concepts. These concepts are extracted via Sparse Autoencoders (SAEs) and subsequently named and annotated on a selected subset of images using a Multimodal LLM. For fair comparison and leakage control, we also introduce the Number of Contributing Concepts (NCC), a decision-level sparsity metric that extends the recently proposed NEC metric. Across diverse datasets, we show that M-CBMs consistently surpass prior CBMs at matched sparsity, while improving concept predictions and providing concise explanations. Our code is available at https://github.com/Antonio-Dee/M-CBM.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2603.07343")
get_code_for_paper("2603.07343")
have("2603.07343")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2603.07343.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2603.07343.svg)](https://syntology.ai/paper/2603.07343)

Connect an agent — have() is free.